GEOPHYSICAL RESEARCH LETTERS, VOL. 37, L07305, doi:10.
1029/2009GL040933, 2010
Click
Here
for
Full
Article
Imaging hydraulic fractures in a geothermal reservoir
Bruce R. Julian,1,2 Gillian R. Foulger,3 Francis C. Monastero,4 and Steven Bjornstad5
Received 30 September 2009; revised 11 January 2010; accepted 18 February 2010; published 8 April 2010.
[1] An injection experiment at the Coso geothermal field in
eastern California in March 2005 caused a swarm of micro-
earthquakes that was recorded by a local network of three‐
component digital seismometers. High‐resolution relative
hypocenter locations propagated upward and northward on
a 700 × 600 m plane striking N 20°E and dipping 75° to
the WNW. This plane is a pre‐existing fault, whose surface
projection coincides with an active scarp. The earthquakes
have similar non‐double‐couple mechanisms that involve
volume increases, and the fault plane bisects their dilata-
tional fields, implying a process dominated by tensile failure.
The source types require the additional involvement of
subsidiary shear faulting, however. Events before and after
the swarm have variable orientations and volume changes
of both signs. Similar tensile‐shear failure is observed in
some natural microearthquake swarms, for example at
Long Valley caldera, California. Its occurrence under low
fluid pressure may imply a heterogeneous stress field or
the induction of thermal stresses by introduction of cold
fluid. Citation: Julian, B. R., G. R. Foulger, F. C. Monastero,
and S. Bjornstad (2010), Imaging hydraulic fractures in a geother-
mal reservoir, Geophys. Res. Lett., 37, L07305, doi:10.1029/
2009GL040933.
1. Introduction
[2] The Coso geothermal area lies in the southwestern
corner of the Basin and Range province in eastern California,
at a right (releasing) step‐over in the southern Owens Valley
fault zone, which experiences 6.5 ± 0.7 mm/year of dextral
shear [Monastero et al., 2005]. The geothermal area has been
exploited since the 1980s to produce electric power.
[3] In February and March of 2005 an “Enhanced
Geothermal Systems” (EGS) experiment was conducted in
injector well 34‐9RD2 on the east flank of the reservoir to
increase permeability and enhance production in a cluster of
wells about 1 km to the south. The well was re‐worked in
order to introduce fluids into a target formation near its
bottom. The existing slotted liner was removed, open frac-
tures and washout regions were cemented and repaired, an
un‐slotted liner was inserted, and the well was then deep-
ened. Major circulation‐loss zones were encountered at a
depth of 2654 m, and a total loss of drilling mud occurred at
1
U.S. Geological Survey, Menlo Park, California, USA.
2
Now at Foulger Consulting, Palo Alto, California, USA.
3
Department of Earth Sciences, Science Laboratories, University of
Durham, Durham, UK.
4
Magma Energy U.S. Corp., Reno, Nevada, USA.
5
Geothermal Program Office, China Lake, California, USA.
Copyright 2010 by the American Geophysical Union.
0094‐8276/10/2009GL040933
L07305
a depth of about 2672 m while injecting water at rates up to
20 l/s.
[4] This injection caused a vigorous swarm of earth-
quakes, which we analyzed using data from a local 36‐station
seismic network. We determined high‐resolution relative
hypocenter locations and complete (symmetric moment‐
tensor) source mechanisms for many of the earthquakes.
Relative hypocenter locations provide information about the
geometry of the failure zone that complements seismic
moment tensors and reduces inherent ambiguities in their
physical interpretation. Surface geological observations
subsequently verified the inferred fault geometry. The
results of this experiment demonstrate that seismological
techniques can provide information of high quality about
hydraulic fractures that are of potential value for operational
decision‐making.
2. Data and Methods
[5] The Geothermal Program Office of the U.S. Navy
monitors seismicity at Coso, operating 22 digital three‐
component short‐period seismometers at depths of about
100 m in custom‐drilled boreholes. To enhance this network
near the planned EGS experiment, we installed 16 additional
digital three‐component seismic stations on the surface. We
choose optimal locations for these stations by computing
theoretical focal‐sphere positions of candidate sites by
numerically tracing rays [Arnott and Foulger, 1994; Julian et
al., 1996; Miller et al., 1998] through a three‐dimensional
crustal wave‐speed model [Wu and Lees, 1999] obtained
from local earthquake tomography. The injection induced a
swarm of about 70 recorded earthquakes, almost all within
the first hour, between 03:00 and 04:00 GMT March 3rd
(Figure 1). Most of the earthquakes occurred in the first two
minutes. A total of 44 earthquakes with M from 0.3 to 2.6
were detected and located by the U.S. Navy permanent
network.
[6] We measured the arrivals from both networks by hand
and located the earthquakes using hypocc, an optimized and
corrected version of the method of Waldhauser and
Ellsworth [2000] to simultaneously invert the inter‐event
arrival‐time differences for many events to obtain accurate
relative hypocenter locations. hypocc speeds computations
by using dynamic storage allocation and efficient algorithms
such as depth‐first graph searching and data structures such
as hash tables [Knuth, 1973, section 6.4] and k‐dimensional
binary search trees [Knuth, 1973, section 6.5] for analyzing
the complex relationships within large data sets.
[7] We computed full (symmetric) moment tensors by
using linear‐programming methods to invert observed P‐,
SH‐and SV‐phase polarities and P:SH, P:SV, and SH:SV
amplitude ratios measured from low‐pass filtered seismo-
grams (corner frequency 5 Hz, 3‐pole Butterworth response),
as described by Julian and Foulger [1996]. Earthquakes
1 of 5
L07305 JULIAN ET AL.: IMAGING HYDRAULIC FRACTURES L07305

L1 norm of the data residuals [Julian and Foulger, 2008]

indicate that the variations found in this study are well re-
solved, as do also the systematic variations of mechanisms
with time discussed below.

Figure 1. Magnitudes of earthquakes within 1.1 km of the
bottom of well 34‐9RD2 for the period 03:00–04:00 March
3rd, 2005.
associated with geothermal areas and hydrofracturing typi-
cally involve significant volume changes, which require
moment tensors for their description [Foulger, 1988;
Foulger and Long, 1984; Foulger and Julian, 1993, 2004;
Foulger et al., 1989; Julian and Foulger, 2004; Julian et al.,
1997, 2004; Miller et al., 1998; Ross et al., 1996, 1999].
[8] The extra degrees of freedom afforded by the moment‐
tensor representation make it important to assess the
uniqueness of derived mechanisms. This is a subject of great
current interest and the state of the art is changing rapidly.
Preliminary confidence regions based on the variation of the
3. Results
[9] The U.S. Navy catalog earthquake hypocenter loca-
tions, which are computed using a one‐dimensional layered
crustal model (K. Richards‐Dinger, personal communica-
tion, 2004), form a diffuse cluster distributed throughout
most of the area of production wells south of the injector
(Figure 2). The accuracy of such hypocenter locations is
difficult to estimate, because the major source of error is
uncertainty about crustal structure, not measurement error.
As is typical for earthquakes located individually in this
manner, the locations do not resolve structures on the scale
of a few tens of meters, as is necessary if the results are to be
of use for scientific understanding and operational decision‐
making. The relative relocations, in contrast, clearly resolve
a plane with a strike of N 20°E dipping 75° to the WNW
with dimensions of about 700–600 m. This activated
structure lies about 500 m southeast of the injector.
[10] We obtained high‐quality moment tensors for 14 of
the largest earthquakes (Figure 3). For comparison, we also
computed moment tensors for seven pre‐injection earth-
quakes and 17 post‐injection earthquakes in March 2005.
The mechanisms of small geothermal earthquakes in tec-
tonic environments typically lie near the line connecting the

Figure 2. Perspective views showing locations of earthquakes: (top) downward‐directed views and (bottom) sub‐horizontal
views with lines of sight along the plane defined by the hypocenters. Blue line: well 34‐9RD2; red lines: wells 34A‐9, 38A‐9,
38C‐9, 38B‐9, 38‐9, 51‐16 and 51A‐16. (left) U.S. Navy catalog locations and (right) relative relocations calculated using
hypocc [Julian and Foulger, 2008]. The green box is 4 × 4 × 4 km. The white lines indicate coordinate directions, and are
centered at sea level directly above the bottom of well 34‐9RD2.
2 of 5
L07305 JULIAN ET AL.: IMAGING HYDRAULIC FRACTURES L07305

Figure 3. (top) Source‐type plots [Hudson et al., 1989] of the mechanisms of (left) pre‐, (middle) co‐ and (right) post‐
injection‐swarm earthquakes. (bottom) Source‐orientation plots showing the principal‐moment directions for the same
earthquake sets. Upper focal hemispheres are shown in equal‐area projection.

+dipole and −dipole points on the source‐type plot of
Hudson et al. [1989] and the post‐injection earthquakes
conform to this pattern. Dipole source types could be
interpreted as opening or closing tensile cracks, with volume
changes partly compensated by fluid inflow. Sources with
implosive components are absent from the pre‐injection
earthquakes, but this absence probably is an artifact of the
small size of the pre‐injection data set; implosive micro-
earthquakes did occur during the previous month. Implosive
mechanisms are entirely lacking from the co‐injection
swarm earthquakes, probably because of either increased
fluid pressure or tensile stresses caused by induced thermal
contraction.
[11] The T axes of most of the earthquakes have
approximately horizontal, east‐west orientations. In contrast,
the P axes show significant variation. For the co‐injection
swarm earthquakes, the T axes are tightly clustered to the
WNW and the P axes occupy a narrow zone extending from
nearly vertical to horizontal and trending to the southwest.
This distribution differs markedly from that either before or
after the injection.
[12] The co‐injection swarm earthquakes have similar
source mechanisms (Figure 4). All the earthquakes studied
have mechanisms consistent with combined normal and
strike‐slip motion — the P polarity plots resemble those of
normal‐faulting earthquakes, but with reduced dilatational
fields with partially explosive mechanisms, i.e., they cor-
respond to opening cavities.
4. Discussion
[13] The interpretation of moment tensors in terms of
physical source processes is not unique [Julian et al., 1998].
The mechanisms shown in Figure 4 could represent shear
slip on faults dipping at various angles to either the WNW
or the ESE, combined with a process such as tensile
cracking that involves a volume increase. Reducing this
ambiguity requires additional independent information
[Foulger et al., 2004].

Figure 4. Moment tensors for 6 typical earthquakes of the 14 studied from the injection swarm, displayed as P‐wave po-
larity plots. Black lines: nodal curves. Red line: the fault delineated by the relative relocations (Figure 2). Open/solid circles:
dilatational/compressional arrivals on the upper focal hemisphere; open/solid squares: dilatational/compressional arrivals on
the lower focal hemisphere; pluses: the center of the focal hemisphere. At right is a schematic illustration of shear wing
faults associated with a propagating tensile crack. The view is along the plane of the crack, which in this case strikes to the
NNE and dips steeply to the WNW.
3 of 5
L07305 JULIAN ET AL.: IMAGING HYDRAULIC FRACTURES L07305

Figure 5. Geological confirmation of the fault delineated by microearthquake hypocenters. (left) Map showing the surface
projection of well 34‐9RD2 (violet), other wells (blue), the surface traces of Quaternary faults (black), and the fault iden-
tified with the earthquake swarm (red). (right) West‐east vertical cross‐section showing well 34‐9RD2 and interpolation
between the surface Quaternary fault scarp and the fault zone imaged in the televiewer log.

[14] On the focal‐sphere plot at the top left of Figure 4,
the fault plane defined by the relative hypocenter locations
(Figure 2) is superimposed on the theoretical P‐phase nodal
curves for the moment tensor. This line, which indicates the
main structure activated, bisects the dilatational field. If the
structure were a shear fault, the line would lie close to a
nodal curve of the focal mechanism, but this is not the case
to a high degree of confidence. A fault bisecting the dila-
tational field is, however, expected for a hydraulically
driven tensile crack [Julian et al., 1998]. Similar observa-
tions have been reported previously, e.g., for naturally
occurring swarm microearthquakes in the south moat of
Long Valley caldera, California [Foulger et al., 2004].
[15] The injection probably stimulated a pre‐existing fault
to fail. The largest earthquake of the swarm was of M 2.6,
which corresponds to failure of a plane with dimensions of
no more than a few tens of metres. This fact suggests that
the ∼600‐m‐long activated fault existed prior to the injection
and was stimulated by the injection to fail in tensile mode.
Each individual earthquake probably represents opening of a
portion of the fault accompanied by subsidiary motion on
shear wing faults (Figure 4, right). Seismic activity propa-
gated northeast and upward during the swarm.
[16] Surface geological observations confirm the existence
of the inferred fault (Figure 5). The scarp of a Quaternary
fault, striking slightly east of north and dipping steeply to the
WNW occurs in surface sediments northeast of the well at
the position obtained by extrapolating the plane of micro-
earthquake hypocenters to the surface. In addition, a tele-
viewer borehole log of well 34‐9RD2 provides evidence of a
fault intersecting the well near its bottom.
[17] Hydraulic fracturing stress tests conducted in nearby
boreholes confirm that the faulting regime of the Coso East
Flank is transitional from normal to strike‐slip. The azimuth
of the smallest horizontal principal stress throughout the
area is 108° ± 24°, a range consistent with the orientation of
the activated fault. The relative magnitudes of ambient
stresses are such that normal faulting can be induced by
increases in reservoir pressure of >3.5 MPa, and strike‐slip
faulting by smaller pressure increases. The magnitude of the
compressive stresses inferred from the hydrofracture tests
are inconsistent with our observation of tensile failure,
however, as the hydraulic pressure of 2672 m of drilling
mud is much smaller (the wellhead pressure was zero). This
inconsistency is presently unresolved. It might point to large
local heterogeneities in stress, to the importance of thermal
stresses caused by the sudden introduction of cold drilling
mud into the fault zone, or some other currently unidentified
process.
5. Conclusions
[18] An injection of drilling mud at a depth of 2672 m in
well 34‐9RD2 at the Coso geothermal area induced a vig-
orous earthquake swarm in March of 2005 that lasted
approximately an hour, with most of the seismic energy
release occurring in the first two minutes. A detailed picture
of fracture formation was obtained from a combination of
relative hypocenter locations and moment tensors.
[19] The swarm activated about 700–600 m of a pre‐
existing fault. This fault opened in tensile mode, with each
small earthquake corresponding to tensile opening of the
main structure, along with subsidiary shear on wing faults
oblique to the main fault. The existence of the structure
deduced from seismic evidence is confirmed by surface
geological observations and by data from a borehole tele-
viewer log. Hydraulic fracturing stress tests indicate omni‐
compressional stresses in boreholes in this part of the
geothermal area, so it is not clear how tensile failure can
occur as a result of the injection of drilling mud under only
hydrostatic pressure. This is a problem associated with
explaining the volumetric components commonly found in
the mechanisms of small earthquakes in fluid reservoirs. It

4 of 5
L07305 JULIAN ET AL.: IMAGING HYDRAULIC FRACTURES
may be an indication that local stresses in such environ-
ments are highly inhomogeneous.
[20] This case history is an important landmark in the
recent development of seismological techniques in support
of Enhanced Geothermal Systems hydrofracturing experi-
ments. It demonstrates that seismological instruments, field
operations, data processing tools and interpretive experience
have matured to the extent that they can deliver information
of utility to operational decision‐making.
[21] Acknowledgments. Funding for this work was provided by
Department of Energy grant DE‐FG3606GO16058 and the U.S. Navy.
We thank Dave Hill and David Shelly for reviews. The data were provided
by the Geothermal Program Office of the U.S. Navy.
References
Arnott, S. K., and G. R. Foulger (1994), The Krafla spreading segment, Ice-
land: 1. Three‐dimensional crustal structure and the spatial and temporal
distribution of local earthquakes, J. Geophys. Res., 99, 23,801–23,825,
doi:10.1029/94JB01465.
Foulger, G. R. (1988), Hengill triple junction, SW Iceland: 2. Anomalous
earthquake focal mechanisms and implications for process within the
geothermal reservoir and at accretionary plate boundaries, J. Geophys.
Res., 93, 13,507–13,523, doi:10.1029/JB093iB11p13507.
Foulger, G. R., and B. R. Julian (1993), Non‐double‐couple earthquakes at
the Hengill‐Grensdalur volcanic complex, Iceland: Are they artifacts of
crustal heterogeneity?, Bull. Seismol. Soc. Am., 83(1), 38–52.
Foulger, G. R., and B. R. Julian (2004), A powerful tool: Use of time‐
dependent MEQ tomography for monitoring producing geothermal reser-
voirs, Geotherm. Res. Counc. Bull., 33, 120–126.
Foulger, G. R., and R. E. Long (1984), Anomalous focal mechanisms:
Tensile crack formation on an accreting plate boundary, Nature, 310
(5972), 43–45, doi:10.1038/310043a0.
Foulger, G. R., R. E. Long, P. Einarsson, and A. Björnsson (1989), Implo-
sive earthquakes at the active accretionary plate boundary in northern
Iceland, Nature, 337(6208), 640–642, doi:10.1038/337640a0.
Foulger, G. R., B. R. Julian, D. P. Hill, A. M. Pitt, P. E. Malin, and E. Shalev
(2004), Non‐double‐couple microearthquakes at Long Valley caldera,
California, provide evidence for hydraulic fracturing, J. Volcanol.
Geotherm. Res., 132(1), 45–71, doi:10.1016/S0377-0273(03)00420-7.
Hudson, J. A., R. G. Pearce, and R. M. Rogers (1989), Source type plot for
inversion of the moment tensor, J. Geophys. Res., 94, 765–774,
doi:10.1029/JB094iB01p00765.
Julian, B. R., and G. R. Foulger (1996), Earthquake mechanisms from
linear‐programming inversion of seismic‐wave amplitude ratios, Bull.
Seismol. Soc. Am., 86(4), 972–980.
Julian, B. R., and G. R. Foulger (2004), Microearthquake focal mechan-
isms: A tool for monitoring geothermal systems, Geotherm. Res. Counc.
Bull., 33, 166–171.
5 of 5
L07305
Julian, B. R., and G. R. Foulger (2008), Moment‐tensor confidence regions
from seismic‐wave amplitude ratios, paper presented at 31st General
Assembly, Eur. Seismol. Comm., Hersonissos, Greece.
Julian, B. R., A. Ross, G. R. Foulger, and J. R. Evans (1996), Three‐
dimensional seismic image of a geothermal reservoir: The Geysers,
California, Geophys. Res. Lett., 23(6), 685–688, doi:10.1029/95GL03321.
Julian, B. R., A. D. Miller, and G. R. Foulger (1997), Non‐double‐couple
earthquake mechanisms at the Hengill‐Grensdalur volcanic complex,
southwest Iceland, Geophys. Res. Lett., 24(7), 743–746, doi:10.1029/
97GL00499.
Julian, B. R., A. D. Miller, and G. R. Foulger (1998), Non‐double‐couple
earthquakes: 1. Theory, Rev. Geophys., 36(4), 525–549, doi:10.1029/
98RG00716.
Julian, B. R., G. R. Foulger, and K. Richards‐Dinger (2004), The Coso
geothermal area: A laboratory for advanced MEQ studies for geothermal
monitoring, paper presented at Annual Meeting, Geotherm. Resour.
Counc., Indian Springs, Calif.
Knuth, D. E. (1973), Sorting and Searching, 723 pp., Addison‐Wesley,
Reading, Mass.
Miller, A. D., B. R. Julian, and G. R. Foulger (1998), Three‐dimensional
seismic structure and moment tensors of non‐double‐couple earthquakes
at the Hengill‐Grensdalur volcanic complex, Iceland, Geophys. J. Int.,
133(2), 309–325, doi:10.1046/j.1365-246X.1998.00492.x.
Monastero, F. C., A. M. Katzenstein, J. S. Miller, J. R. Unruh, M. C.
Adams, and K. Richards‐Dinger (2005), The Coso geothermal field: A
nascent metamorphic core complex, Geol. Soc. Am. Bull., 117(11–12),
1534–1553, doi:10.1130/B25600.1.
Ross, A., G. R. Foulger, and B. R. Julian (1996), Non‐double‐couple earth-
quake mechanisms at the Geysers geothermal area, California, Geophys.
Res. Lett., 23(8), 877–880, doi:10.1029/96GL00590.
Ross, A., G. R. Foulger, and B. R. Julian (1999), Source processes of
industrially‐induced earthquakes at the Geysers geothermal area,
California, Geophysics, 64(6), 1877–1889, doi:10.1190/1.1444694.
Waldhauser, F., and W. L. Ellsworth (2000), A double‐difference earth-
quake location algorithm: Method and application to the northern
Hayward fault, California, Bull. Seismol. Soc. Am., 90(6), 1353–1368,
doi:10.1785/0120000006.
Wu, H., and J. Lees (1999), Three‐dimensional P and S wave velocity
structures of the Coso geothermal area, California, from microseismic
travel time data, J. Geophys. Res., 104, 13,217–13,233, doi:10.1029/
1998JB900101.
S. Bjornstad, Geothermal Program Office, 429 East Bowen Rd., Stop
4011, China Lake, CA 93555‐6108, USA.
G. R. Foulger, Department of Earth Sciences, Science Laboratories,
University of Durham, South Road, Durham DH1 3LE, UK.
B. R. Julian, Foulger Consulting, 1025 Paradise Way, Palo Alto, CA
94306, USA. (bruce@foulgerconsulting.com)
F. C. Monastero, Magma Energy U.S. Corp., 9740 S. McCarran Blvd.,
Ste. 103, Reno, NV 89523‐9204, USA.